65P - Multiplex immunofluorescence analysis of LRRC15 and the TME in early-stage lung adenocarcinoma

Background

This study describes a six-marker multiplex immunofluorescence (mIF) panel, along with novel marker, LRRC15. Expression in cancer is typically in the stromal compartment. However, in cancers of mesenchymal origin, LRRC15 was also found on tumour cells. Recent studies suggest roles for LRRC15 in invasion and immune modulation. This study demonstrates that mIF paired with machine learning significantly advances our ability to classify and analyse tissue samples from cancer patients.

Methods

A tissue microarray (TMA) of 174 cases of early-stage lung adenocarcinoma with 8 TMA cores from each patient was used for mIF. Tumour cells were distinguished from stromal cells with pan-cytokeratin, while markers for CD68, CD3, αSMA, vimentin, and LRRC15 were used to study immune infiltrates and the stromal compartment. Indica HALO AI image analysis platform was used to classify and analyse mIF images. Density and population of multiple cell phenotypes were then calculated. Classification models were trained and the Kruskal Wallis algorithm was used to rank importance of phenotypes. Kaplan-Meier survival curves were then plotted for highest ranked phenotypes.

Results

Several phenotypes displayed predictors of survival that outperformed previous prognostic scores. Top phenotypes from Kaplan-Meier survival analysis showed that in tumour area, high LRRC15 density predicts poorer 5-year survival in patients (HR: 1.61, 95% CI: 0.956 to 2.72, P=0.044), while high CD68 density predicts better 5-year survival (HR: 0.472, 95% CI: 0.296 to 0.753, P= 0.0006). Furthermore, high CD68 density with LRRC15 exclusion is a more powerful predictor of 5-year survival (HR: 0.444, 95% CI: 0.279 to 0.708, P= 0.0002).

Conclusions

We have demonstrated a mIF and machine learning pipeline that could enhance the performance of survival predictors. Furthermore, understanding LRRC15 in the TME could contribute to precision medicine in lung cancer.

Editorial acknowledgement

Clinical trial identification

Legal entity responsible for the study

University of St Andrews.

Funding

Melville Trust.

Disclosure

All authors have declared no conflicts of interest.